pdf-inspector ships two command-line tools with the crate — cargo install pdf-inspector installs both in one go, no code required for classification or Markdown conversion. The npm package bundles the CLI too.

Install

# Rust toolchain users
cargo install pdf-inspector

# Or Node users (CLI bundled in the npm package)
npm install -g @firecrawl/pdf-inspector

You get two executables:

Tool	What it does	Typical time
detect-pdf	Classify the document + analysis info	~10–50ms
pdf2md	Full parse; outputs Markdown	<200ms for text-based

detect-pdf: ask "can I extract this directly?" first

# Quick verdict
detect-pdf document.pdf

# With analysis info + JSON output (recommended)
detect-pdf document.pdf --analyze --json

The most useful JSON fields:

{
  "pdfType": "TextBased",
  "confidence": 0.92,
  "pagesNeedingOcr": []
}

TextBased → run pdf2md with confidence. Scanned/Mixed → you know exactly which pages need OCR. This classify-first routing is the officially recommended way to keep costs down.

pdf2md: one command to Markdown

# Print to terminal
pdf2md annual-report.pdf

# Write to a file
pdf2md annual-report.pdf -o report.md

# Selected pages only
pdf2md annual-report.pdf --pages 1-3

# Structured JSON (metadata included, easy for programs)
pdf2md annual-report.pdf --json

# Token-saving compact output (before feeding LLMs)
pdf2md annual-report.pdf --compact

Flags at a glance

Usage	Meaning
pdf2md <file>	Parse; Markdown goes to stdout
pdf2md <file> -o out.md	Write to a file
pdf2md <file> --pages 1-3	Process selected pages only
pdf2md <file> --select-pages	Interactively select pages
pdf2md <file> --compact	Compact, token-saving output
pdf2md <file> --json	Structured JSON output
detect-pdf <file>	Quick classification
detect-pdf <file> --analyze --json	Classification + analysis

Common combinations

Batch-convert a directory

for f in *.pdf; do
  pdf2md "$f" -o "${f%.pdf}.md"
done

A classify-first routing script

type=$(detect-pdf "$f" --analyze --json | jq -r '.pdfType')
if [ "$type" = "TextBased" ]; then
  pdf2md "$f" -o "${f%.pdf}.md"
else
  echo "Skipping $f: $type needs OCR"
fi

Pipe it onward

pdf2md annual-report.pdf --compact | your-llm-loader

Tips

• The first cargo install compiles for a few minutes; after that everything runs at native speed
• Prebuilt binaries exist for Windows, macOS ARM, and Linux x64/ARM64 — the npm route skips compilation entirely
• For every flag, trust the repo README and --help